Signal Over Noise #33
December 17th, 2025
Dear Reader,
Last week I asked Claude a simple question: “What patterns do you see in how I’ve been working this month?”
Some hours later, I was reviewing a complete redesign of how I approach my projects.
Here’s the thing — I’d been using AI to do things all month: writing code, researching topics, generating content. The usual productivity stuff. But I realised that I was in a perfect spot to ask it to help me think about what I was doing, and asking that question has changed how I work with AI since.
The Problem with How Most of Us Use AI
Most AI conversations look like this:
- “Write me a…”
- “Create a…”
- “Help me with…”
Task, output, done — move on to the next thing.
We’re using these tools like very smart search engines or fancy autocomplete. Nothing wrong with that — it’s useful. But there’s a whole other way to use them that most people never try: asking questions that help you work better, not just faster.
How I Set This Up (The Simple Version)
My setup now largely consists of running Claude Code on the left, and Obsidian on the right.
I’m going to be honest - I have a fairly elaborate system. Claude Code reads my daily notes, I have automated summaries, weekly review processes, the whole thing.
But here’s what actually matters: the habit of capturing and the questions you ask. If you’ve ever had a journal (or frankly, taken meeting notes), you know the drill.
But you don’t need my infrastructure. Here’s the AI-agnostic version:
Step 1: Capture Something (Anywhere)
Pick one place — Google Doc, Apple Notes, Notion, even a plain text file. The tool doesn’t matter.
Each day, spend 2 minutes jotting down:
- What you worked on
- What felt hard
- What felt easy
- Any decisions you made
That’s it. Not a journal or detailed notes — just breadcrumbs for your future self.
Step 2: Weekly Review (Two Prompts)
At the end of the week, paste your notes into ChatGPT and start with:
“Based on these notes, give me a summary of what I’ve been working on this week. What were my main focuses? What took most of my time?”
This does two things: it forces the AI to actually understand your context, and it lets you verify it “got it right” before going deeper. You’ll often see your week differently when it’s reflected back to you.
Then follow up with:
“What patterns do you see? What am I spending energy on that might not be serving me? What should I do more of?”
The AI doesn’t know you, but that’s actually the point — it can spot patterns in your notes that you’re too close to see yourself.
Step 3: Year-End Reflection (What We’re Doing Today)
Same idea, bigger timeframe. Paste in your December notes — or even just key highlights you remember from the year — and ask the questions below.
The Questions Worth Asking
These are the exact prompts I’ve been using. They work in ChatGPT, Claude, Gemini — any of them (but my top tip is to use a reasoning model, like Opus 4.5, instead of a writing/personality-driven model).
For Reviewing How You Work
“Based on [this context], what am I over-complicating?”
I asked this about my December work. Claude pointed out I’d spent 8 hours on a “bootstrap system” to do something that could be done with a checklist. Ouch. True.
“What problems am I solving repeatedly that I should automate or eliminate?”
This one surfaced that I was manually doing the same research process every week when a simple template would cut it in half.
“Where am I confusing activity with progress?”
Brutal, but necessary, especially if you like to tune and tweak systems.
For Planning What’s Next
“Based on what’s worked this month, what should I prioritize in January?”
Not “what do I want to do” — what does the evidence suggest?
“What’s one thing I should stop doing that I probably won’t?”
I love this question because it forces honesty about habits vs. intentions.
“If I could only work on three things next month, what would have the biggest impact?”
Constraint forces clarity.
For Skills and Learning
“What skills am I using most? What skills am I avoiding?”
My December answer: heavy on prompt engineering and automation, conspicuously light on video and public speaking. That pattern tells me something.
“Based on how I work, what should I learn next that would multiply my effectiveness?”
Better than “what’s trendy to learn.”
What I Actually Discovered
When I ran this process on my December work, here’s what emerged:
- I tend to over-engineer Built a 22-repository bootstrap system when 5 would have covered 90% of use cases. The AI called it out clearly: “You’re spending time on infrastructure that might not be needed.”
- Simple prompts work better. I’d been crafting elaborate, structured prompts. My best results came from direct questions. The complexity was for me, not for the AI.
- I don’t review enough. Lots of building, not enough stepping back. The pattern was obvious in hindsight.
None of this was hidden. It was all in my notes. I just needed something to point at it and say “hey, notice this?”
The Questions You Should Try This Week
Pick one. Just one. Here’s my recommendation for where to start:
If you have notes/context to share: Paste a month of work into your AI of choice and ask: “What patterns do you see in how I work? What should I be thinking about as I plan next year?”
If you’re starting from scratch: Open ChatGPT and type: “I want to reflect on how I’ve been working this year. Ask me 5 questions that would help me think about what to change in 2025.”
Then actually answer them — out loud or in text — and let the AI follow up.
If you want to go deeper: At the end of the conversation, ask: “Based on everything I’ve shared, what’s the one thing you’d tell me that I probably don’t want to hear?”
That last one is where the value lives.
One more thing to try: This weekend I took summaries of my “want to learn” list from Domestika and LinkedIn Learning and asked Claude to help me develop a learning plan that complements my skills, and my objectives. It chucked away a lot of cruft on topics that I already knew, and things that aren’t “mission critical” (with the exception of some hobby-driven things). Sometimes just using an AI agent as a second pair of eyes helps strip away our own complications.
Making This a Habit (Not Just a December Thing)
The year-end review is valuable, but the real payoff comes from doing this regularly.
Here’s the minimum viable version:
- Daily: 2 minutes of notes (what you did, what was hard)
- Weekly: Paste notes into AI, ask for a summary first, then “what patterns do you see?”
- Monthly: Ask “what should I prioritize next month based on this month?”
You don’t need fancy tools. You need the habit and the questions.
Your Turn
Try one of these prompts this week. Not because I said so — because December is actually a good time to think about how you’re working before the new year starts.
Hit reply and tell me what you discovered. I’m genuinely curious what patterns show up for others.
Until next week,
Jim
€9.00
The AI Writing Field Guide
Stop Sounding Like Everyone Else Using AI
The 5-step system for teaching AI what your voice actually sounds like — not… Read more
Signal Over Noise is weekly, reader-first publication on AI “without the hype” published by Jim Christian. If you’ve been forwarded this issue, you can subscribe for free: go.signalovernoise.at.
Made with ❤️ in Valencia by Jim Christian. For feedback, please reach out to [email protected].
